Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!lll-winken!uunet!mcvax!ukc!strath-cs!nott-cs!ucl-cs!A.Balou From: A.Balou@ucl-cs.UUCP Newsgroups: comp.lang.smalltalk Subject: Intra-object Parallelism Message-ID: <219@ucl-cs.UUCP> Date: 11 Apr 89 15:10:52 GMT Sender: A.Balou@ucl-cs.UUCP Lines: 40 From: Angeliki One of the levels O-O parallelism can be effected, is at method-level (what I call "intra-object" parallelism). Of course, there are many ways to do that, with one extreme having "pure" atomic objects (each method runs to completion) and the other having fully "re-entrant" objects (all messages activate methods and synchronization is achieved from within the methods). Many different "flavours" cover the spectrum between the extremes. A number of issues is related to "intra-object" parallelism, like the sort of synchronization mechanism, who is in control of the parallelism etc To my view the kind of parallelism depends : - on the intended applications - on the language level (mine is intermediate-level) Besides, I believe that any kind of "intra-object" parallelism (or none at all) can be expressed in terms of the other(s) with more or less effort. Nevertheless, as I find this issue very interesting (also in terms of object modelling), I would like to discuss it with those who share my interest ... PS. Please, don't be too fussy with my terminology: I 'm just trying to avoid repetition of wordy definitions. So, by "atomic" objects, I imply objects which allow at most one method being active at any moment, whereas "re-entrant" are the objects allowing concurrent execution of more than one of their methods. ----------------------------------------------------------------- Angeliki Balou Dept. of Computer Science University College London Gower Street London WC1E 6BT UK e-mail address: angeliki@cs.ucl.ac.uk ------------------------------------------------------------------